隐写分析技术
人工智能
计算机科学
模式识别(心理学)
特征提取
残余物
判别式
深度学习
卷积神经网络
卷积(计算机科学)
图像融合
人工神经网络
特征(语言学)
隐写术
图像(数学)
算法
语言学
哲学
作者
Chengliang Xie,Xiangjun Wu
摘要
Deep steganalysis networks (DSNs) has made a great progress in detection performance. However, most of the deep image steganalysis networks are not complete end-to-end models. In their approach, traditional hand-crafted features are employed to pre-process the images, which can obtain the high-frequency noise residuals to alleviate the interference of the image content. To avoid relying on domain knowledge of deep learning-based methods, we design an end-to-end deep image steganalysis neural model that combines multi-scale feature extraction and residual fusion modules. Firstly, we use the standard convolution kernels of different sizes to extract the features of different scale receptive fields in the input image. Then, depth-wise convolution is used to independently models the inner-channel correlations of multi-scale features and retains the discriminative statistical characteristics of each channel. The residual fusion technique is introduced to aggregate hierarchical feature and strengthen information representation in the network. Experimental results show that, compared with the existing classical deep image steganalysis networks, the proposed steganalysis scheme has a great improvement in steganalysis error rates.
科研通智能强力驱动
Strongly Powered by AbleSci AI